Continuity of stochastic integrals with respect to infinitely divisible random measures, (with J. Rosinski).

dvi (164K)

Abstract:

Sufficient conditions for boundedness and continuity are obtained for stochastically continuous infinitely divisible processes, without Gaussian component, $ \{Y(t), t\in T\}$, where $T$ is a compact metric space or pseudo-metric space. Such processes have a version given by $ Y(t)=X(t)+b(t)$, $t\in T$ where $b$ is a deterministic drift function and $$ X(t)= \int_S f(t,s)\Big[N(ds)-(|f(t,s)|\vee 1)^{-1}\, \nu(ds)\Big]. $$ Here $N$ is a Poisson random measure on a Borel space $S$ with $\sigma$--finite mean measure $\nu$, and $f: T\times S\mapsto \R$ is a measurable deterministic function.


Sufficient conditions for continuity and boundedness of moving average processes, (with J. Rosinski) dvi (57K)

Abstract:

We show that the moving average process \[ {\mathcal X}_f(t):= \int_{0}^t f(t-s) \, dZ(s) \qquad t \in [0,T] \] has a bounded version almost surely, when the kernel $f$ has finite total 2--variation and $Z$ is a symmetric L\'evy process. We also obtain bounds for $E|\sup_{t\in[0,t]}{\mathcal X}_f(t)|$ and for uniform moduli of continuity of ${\mathcal X}_f(\cd)$ and for the largest jump of ${\mathcal X}_f(t)$ when it is not continuous. Similar results are obtained for forward average processes. The methods developed are also used to show that certain infinitely divisible random fields are bounded.


Local times on curves dvi (43K)

Abstract:

Given a positive measure $\mu$ on $R^n$, a continuous additive functional of an $R^n$ valued Levy process $X$ can be defined (heuristically) by \[ L^{\mu}_{t}=\lim_{\ep\to 0}\int \int_0^tf_\ep(X(s)-x)\,ds\,d\mu(x) \] where $f_\ep$ is an approximate $\de$ function at zero. If the support of $\mu$ is a curve in $R^2$ or a surface in $R^n$ we can consider $L_t^\mu$ as the local time on the curve or surface. Given a family $\MM$ of such measures, endowed with a metric, some sufficient conditions for the continuity of $\{L_t^\mu,(t,\mu)\in R_+\times\MM\}$ are obtained. This problem is related, in part, to the continuity of certain second order Gaussian chaos processes indexed by the functions in $\MM$.


Two results on the continuity and boundedness of stochastic convolutions, (with S. Kwapien and J. Rosinski) dvi ( K)

Abstract:

Two results are given concerning the continuity and boundedness of stochastic convolutions of the form $$ {\mathcal X}_f(t) := \int_{0}^t f(t-s) \, dZ(s) \qquad t \in [0,1] $$ where $Z$ is a symmetric L\'evy process and $f: [0,1] \mapsto R$ is a continuous function with $f(0)=0$. The first is that the process ${\mathcal X}_f(t)$ is contiunous almost surely when $f$ is a sample path of any continuous Gaussian process with stationary increments, which is independent of $Z$. This shows, in particular, that there are continuous moving averages for which $f$ has arbitrarily large modulus of continuity. The second goes in the opposite direction. It shows that for any process $Z$, with paths of infinite variation, there exists a continuous function $f$, with $f(0)=0$, such that ${\mathcal X}_f(t)$ has unbounded paths almost surely.